Why is IceCat now only 'supported' on Intel-systems?

Why is IceCat now only 'supported' on Intel-systems?

[Mark H Weaver]

According to:

  https://hydra.gnu.org/eval/109381?filter=icecat#tabs-removed

the jobs for icecat on armhf and mips64el were removed in evaluation
109381 (corresponding to commit 663d5b5), but were present in the
previous evaluation 109380 (commit cd65d60).

Can anyone tell me why this happened?

I guess that the 'supported-systems' field of some package that 'icecat'
depends on was recently changed, but I was unable to find anything
obvious by grepping through the output of "git log -p".

Debian includes Firefox packages for 'armhf' and 'mips64el', so it's
obviously possible to get it working on those platforms.

I find it disturbing that we seem to be in the habit of removing
non-Intel systems from 'supported-systems' fields in packages that other
distros are able to get working on non-Intel.  These are bugs to be
fixed, not swept under the rug to get them out of sight.

      Mark

Re: Why is IceCat now only 'supported' on Intel-systems?

[Tobias Geerinckx-Rice]

[-- Attachment #1.1: Type: text/plain, Size: 1235 bytes --]

Mark

On 01/12/16 03:39, Mark H Weaver wrote:
> According to:
> 
>   https://hydra.gnu.org/eval/109381?filter=icecat#tabs-removed
> 
> the jobs for icecat on armhf and mips64el were removed in evaluation
> 109381 (corresponding to commit 663d5b5), but were present in the
> previous evaluation 109380 (commit cd65d60).
> 
> Can anyone tell me why this happened?

Most very likely my 7cba764 commit, ‘icecat: Use libjpeg-turbo instead
of bundled libjpeg’[-turbo].

The ‘libjpeg-turbo’ package depends on ‘nasm’, which I didn't realise
was only supported on x86.

> Debian includes Firefox packages for 'armhf' and 'mips64el', so it's
> obviously possible to get it working on those platforms.

I'll take a look. IceCat's bundled ‘libjpeg’ is really the (a?) -turbo
variant, so it shouldn't matter, but I obviously missed something.

> I find it disturbing that we seem to be in the habit of removing
> non-Intel systems from 'supported-systems' fields in packages that other
> distros are able to get working on non-Intel.  These are bugs to be
> fixed, not swept under the rug to get them out of sight.

While a valid sentiment, that's not what happened here.

Kind regards,

T G-R


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 476 bytes --]

Re: Why is IceCat now only 'supported' on Intel-systems?

[Tobias Geerinckx-Rice]

[-- Attachment #1.1: Type: text/plain, Size: 315 bytes --]

On 01/12/16 15:13, Tobias Geerinckx-Rice wrote:
> Most very likely my 7cba764 commit, ‘icecat: Use libjpeg-turbo instead
> of bundled libjpeg’[-turbo].

After some more poking around Hydra I might have been too eager to take
the blame for this. I'll keep an eye on it, though.

Kind regards,

T G-R


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 476 bytes --]

Re: Why is IceCat now only 'supported' on Intel-systems?

[Ludovic Courtès]

Mark H Weaver <mhw@netris.org> skribis:

> According to:
>
>   https://hydra.gnu.org/eval/109381?filter=icecat#tabs-removed
>
> the jobs for icecat on armhf and mips64el were removed in evaluation
> 109381 (corresponding to commit 663d5b5), but were present in the
> previous evaluation 109380 (commit cd65d60).
>
> Can anyone tell me why this happened?

“git diff 663d5b5..cd65d60 | grep supported-sy” shows nothing.
I don’t see anything obvious in the logs.

Indeed:

--8<---------------cut here---------------start------------->8---
scheme@(guile-user)> ,use(guix)
scheme@(guile-user)> ,use(gnu packages gnuzilla)
scheme@(guile-user)> (package-supported-systems icecat)
$2 = ("x86_64-linux" "i686-linux" "armhf-linux" "mips64el-linux")
scheme@(guile-user)> (system* "git" "describe")
v0.11.0-3128-gd30e578
$3 = 0
--8<---------------cut here---------------end--------------->8---

In the current evaluation IceCat is there on all arches, though it only
succeeds on x86_64:

  https://hydra.gnu.org/eval/109382?filter=icecat&compare=109381&full=#tabs-now-succeed

> I find it disturbing that we seem to be in the habit of removing
> non-Intel systems from 'supported-systems' fields in packages that other
> distros are able to get working on non-Intel.  These are bugs to be
> fixed, not swept under the rug to get them out of sight.

Everyone will agree with that, and everyone will agree that Someone must
fixed said bugs.

There’s also a practical issue not to be swept under the rug: when we
keep rebuilding known-to-fail packages on those architectures, we waste
build resources.

Ludo’.

Re: Why is IceCat now only 'supported' on Intel-systems?

[Efraim Flashner]

[-- Attachment #1: Type: text/plain, Size: 1761 bytes --]

On Wed, Nov 30, 2016 at 09:39:40PM -0500, Mark H Weaver wrote:
> According to:
> 
>   https://hydra.gnu.org/eval/109381?filter=icecat#tabs-removed
> 
> the jobs for icecat on armhf and mips64el were removed in evaluation
> 109381 (corresponding to commit 663d5b5), but were present in the
> previous evaluation 109380 (commit cd65d60).
> 
> Can anyone tell me why this happened?
> 
> I guess that the 'supported-systems' field of some package that 'icecat'
> depends on was recently changed, but I was unable to find anything
> obvious by grepping through the output of "git log -p".
> 
> Debian includes Firefox packages for 'armhf' and 'mips64el', so it's
> obviously possible to get it working on those platforms.
> 
> I find it disturbing that we seem to be in the habit of removing
> non-Intel systems from 'supported-systems' fields in packages that other
> distros are able to get working on non-Intel.  These are bugs to be
> fixed, not swept under the rug to get them out of sight.
> 
>       Mark

It turns out libjpeg-turbo depends on nasm, which was incorrectly marked
as only supporting Intel hardware. I think the hard thing is that, while
we can read the source and make guesses at fixing build errors, without
access to the hardware its hard to know if any changes made actually
result in a building package.

Anyway, Debian builds nasm on all architectures, so I removed that field
from the package. Looking back on the nasm thread, it looked like
everyone worked together to make the mistake :).

-- 
Efraim Flashner   <efraim@flashner.co.il>   אפרים פלשנר
GPG key = A28B F40C 3E55 1372 662D  14F7 41AA E7DC CA3D 8351
Confidentiality cannot be guaranteed on emails sent or received unencrypted

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: Why is IceCat now only 'supported' on Intel-systems?

[Ludovic Courtès]

Efraim Flashner <efraim@flashner.co.il> skribis:

> On Wed, Nov 30, 2016 at 09:39:40PM -0500, Mark H Weaver wrote:
>> According to:
>> 
>>   https://hydra.gnu.org/eval/109381?filter=icecat#tabs-removed
>> 
>> the jobs for icecat on armhf and mips64el were removed in evaluation
>> 109381 (corresponding to commit 663d5b5), but were present in the
>> previous evaluation 109380 (commit cd65d60).
>> 
>> Can anyone tell me why this happened?
>> 
>> I guess that the 'supported-systems' field of some package that 'icecat'
>> depends on was recently changed, but I was unable to find anything
>> obvious by grepping through the output of "git log -p".
>> 
>> Debian includes Firefox packages for 'armhf' and 'mips64el', so it's
>> obviously possible to get it working on those platforms.
>> 
>> I find it disturbing that we seem to be in the habit of removing
>> non-Intel systems from 'supported-systems' fields in packages that other
>> distros are able to get working on non-Intel.  These are bugs to be
>> fixed, not swept under the rug to get them out of sight.
>> 
>>       Mark
>
> It turns out libjpeg-turbo depends on nasm, which was incorrectly marked
> as only supporting Intel hardware.

Again, there’s no such problem on current master AFAICS:

--8<---------------cut here---------------start------------->8---
scheme@(guile-user)> ,use(guix)
scheme@(guile-user)> ,use(gnu packages assembly)
scheme@(guile-user)> ,use(gnu packages gnuzilla)
scheme@(guile-user)> (package-supported-systems icecat)
$15 = ("x86_64-linux" "i686-linux" "armhf-linux" "mips64el-linux")
scheme@(guile-user)> (package-supported-systems nasm)
$16 = ("x86_64-linux" "i686-linux" "armhf-linux" "mips64el-linux")
scheme@(guile-user)> (package-transitive-supported-systems icecat)
$17 = ("x86_64-linux" "i686-linux" "armhf-linux" "mips64el-linux")
--8<---------------cut here---------------end--------------->8---

and <https://hydra.gnu.org/eval/109381?filter=icecat>.

Am I missing something?

Ludo’.

Re: Why is IceCat now only 'supported' on Intel-systems?

[Mark H Weaver]

ludo@gnu.org (Ludovic Courtès) writes:

> Efraim Flashner <efraim@flashner.co.il> skribis:
>
>> On Wed, Nov 30, 2016 at 09:39:40PM -0500, Mark H Weaver wrote:
>>> According to:
>>> 
>>>   https://hydra.gnu.org/eval/109381?filter=icecat#tabs-removed
>>> 
>>> the jobs for icecat on armhf and mips64el were removed in evaluation
>>> 109381 (corresponding to commit 663d5b5), but were present in the
>>> previous evaluation 109380 (commit cd65d60).
>>> 
>>> Can anyone tell me why this happened?
>>> 
>>> I guess that the 'supported-systems' field of some package that 'icecat'
>>> depends on was recently changed, but I was unable to find anything
>>> obvious by grepping through the output of "git log -p".
>>> 
>>> Debian includes Firefox packages for 'armhf' and 'mips64el', so it's
>>> obviously possible to get it working on those platforms.
>>> 
>>> I find it disturbing that we seem to be in the habit of removing
>>> non-Intel systems from 'supported-systems' fields in packages that other
>>> distros are able to get working on non-Intel.  These are bugs to be
>>> fixed, not swept under the rug to get them out of sight.
>>> 
>>>       Mark
>>
>> It turns out libjpeg-turbo depends on nasm, which was incorrectly marked
>> as only supporting Intel hardware.
>
> Again, there’s no such problem on current master AFAICS:
>
> --8<---------------cut here---------------start------------->8---
> scheme@(guile-user)> ,use(guix)
> scheme@(guile-user)> ,use(gnu packages assembly)
> scheme@(guile-user)> ,use(gnu packages gnuzilla)
> scheme@(guile-user)> (package-supported-systems icecat)
> $15 = ("x86_64-linux" "i686-linux" "armhf-linux" "mips64el-linux")
> scheme@(guile-user)> (package-supported-systems nasm)
> $16 = ("x86_64-linux" "i686-linux" "armhf-linux" "mips64el-linux")
> scheme@(guile-user)> (package-transitive-supported-systems icecat)
> $17 = ("x86_64-linux" "i686-linux" "armhf-linux" "mips64el-linux")
> --8<---------------cut here---------------end--------------->8---

I believe that Efraim fixed the issue in commit
4181c9393d46bb8e24af1970c0e2adbcbfb2c639, shortly after my initial post
in this thread.

> and <https://hydra.gnu.org/eval/109381?filter=icecat>.

The URL above shows that the icecat jobs for armhf and mips64el were
removed, but that was before Efraim's fix.

     Thanks,
       Mark